Re: [PATCH 1/2] staging: r8188eu: Use usb_control_msg_recv/send() in usbctrl_vendorreq()

[Pavel Skripkin]

On 8/24/21 11:53 AM, Fabio M. De Francesco wrote:
> On Tuesday, August 24, 2021 10:13:46 AM CEST Pavel Skripkin wrote:
> > On 8/24/21 1:37 AM, Fabio M. De Francesco wrote:
> > > Replace usb_control_msg() with the new usb_control_msg_recv() and
> > > usb_control_msg_send() API of USB Core in usbctrl_vendorreq().
> > > 
> > > Suggested-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@xxxxxxxxx>
> > > ---
> > > 
> > > Thanks to Pavel Skripkin <paskripkin@xxxxxxxxx> for his review of the
> > > RFC patch.
> > >   
> > > drivers/staging/r8188eu/hal/usb_ops_linux.c | 25 ++++++++++-----------
> > > 1 file changed, 12 insertions(+), 13 deletions(-)
> > > 
> > > [...]
> > >
> > Hi, Fabio!
> >
> > Christophe is right about semantic part. 
>
> Hi Pavel,
>
> I haven't yet read Christophe's message (but I'm going to do it ASAP).
> I hope he found out what is wrong with the code, what made Phil's tests
> fail.
>
> > Also,
> >
> > if (!status) {
> >
> > } else {
> > 	if (status < 0) {		<-
> > 					  |
> > 	} else {			  |
> > 					  |
> > 	}				<-
> > }					
> >
> > Extra if-else is not needed, since status can be 0 and < 0, there is no 
> > 3rd state, like it was before.
>
> Correct, thanks!
>
> Now I read the following from the documentation of the new API...
>
> "Return: If successful, 0 is returned, Otherwise, a negative error number."
>
> I'll remove that status < 0 check and whatever else is no more necessary.
> Thanks, again :)
>
> Regards,

Btw, not related to your patch, but I start think, that this check:


	if (!pIo_buf) {
		DBG_88E("[%s] pIo_buf == NULL\n", __func__);
		status = -ENOMEM;
		goto release_mutex;
	}

Should be wrapped as

	if (WARN_ON(unlikely(!pIo_buf)) {
		...
	}

Since usb_vendor_req_buf is initialized in ->probe() and I can't see 
possible calltrace, which can cause zeroing this pointer.

Something _completely_ wrong is going on if usb_vendor_req_buf is NULL, 
and we should complain loud about it. What do you think?


With regards,
Pavel Skripkin